[an error occurred while processing this directive] [an error occurred while processing this directive][an error occurred while processing this directive] [an error occurred while processing this directive] [an error occurred while processing this directive] [an error occurred while processing this directive] (none) [an error occurred while processing this directive] [an error occurred while processing this directive] [an error occurred while processing this directive] [an error occurred while processing this directive] [an error occurred while processing this directive][an error occurred while processing this directive] [an error occurred while processing this directive][an error occurred while processing this directive] [an error occurred while processing this directive][an error occurred while processing this directive] [an error occurred while processing this directive] [an error occurred while processing this directive] [an error occurred while processing this directive] (none) [an error occurred while processing this directive] [an error occurred while processing this directive] [an error occurred while processing this directive][an error occurred while processing this directive]
 
[an error occurred while processing this directive] [an error occurred while processing this directive]
Skåne Sjælland Linux User Group - http://www.sslug.dk Home   Subscribe   Mail Archive   Forum   Calendar   Search
MhonArc Date: [Date Prev] [Date Index] [Date Next]   Thread: [Date Prev] [Thread Index] [Date Next]   MhonArc
 

Re: [SCICOMP] Announce: TONS-0.0.1 released



sslug@sslug wrote:
> 
> To whom it may concern:
> 
> The TONS project (http://www.sslug.dk/TONS/) aims to produce an
> environment in which scientists and engineers can accomplish numerical
> calculations, easier and faster than before.
> 
> The 0.0.1 release is barely a proof-of-concept release.  In fact, it
> includes nothing of real value to end users.  The code is released only
> to fuel interest in the project, in the hope that more developers and
> clever people in general will find interest in the project.
> 
> Goals:
> 
>  *)  To produce a computing environment that will provide an alternative to
>      MatLab, Octave, RLab, and the others (maybe even C/Fortran + MPI).
>      The environment will have to consist of a good language, visualization
>      capabilities, and an effective computing engine.
> 
>  *)  TONS will automatically parallelize code. This will make it possible to
>      write software _once_ and run on any cluster configuration without new
>      tuning or rewriting of the code.
> 
> Current status:
> 
>  *)  It compiles.    :)
> 
>  *)  Basically everything needs to be done. Sample code exists for computation
>      of Fibonacci numbers, and this can be run over a network (but the code is
>      not yet parallelized).
> 
> The future:
> 
>  *)  A compiler is in the works.
>      Design of a language is something that should not be done over night.  And
>      the design of the TONS Algorithmic Language (TAL) is bound to be an effort
>      that will take time.
> 
>  *)  A lot of effort is currently being put into the automatic parallelization.
>      This feature of TONS will hopefully evolve rapidly in the near future.
>      The goal is to have good automatic parallelization running in early July.
> 
>  *)  Expect to see frequent releases, as the mist clears.

Det lyder som et rigtig godt projekt. Hvis bare man havde tid til alt
det man gerne ville lave...

Jeg kan se der er sket en del med C++, siden jeg sidst rigtig
beskæftigede mig med det, "namespaces"???

Er det smart at alt kode bliver udført som byte-kode?? I stor-skala
beregninger higer man vel efter så meget kraft så muligt, og selvom TASM
bytekoden nok kan laves meget mere effektivt en fx. JVM bytekode, så
skal det stadigvæk fortolkes. Måske en mulighed for at køre "native"
kode, hvis man har en homogen cluster kunne tilføjes. 

Såvidt jeg kan se er der ikke lavet meget på TAL. Automatisk
parallisering kan vel nærmest betegnes som den hellige gral inden for
parallelprogrammering i øjeblikket, og det kunne være ekstremt køligt at
have noget sådant. 

Jeg kunne umiddelbart se en fordel i en explicit parallelisering
indbygget i programmeringssproget, måske på en lidt nemmere måde end med
trådabstraktionen?

Fx. som Hoare (eller var det Dijkstra?) lavede:

...
cobegin
	.. 
	gang nogle matricer sammen
[]
	gang nogle andre matricer sammen
	..
coend

Jeg kunne også se det smarte i at have end tilsvarende
distbegin...distend, hvor "dist" står for "distribueret". Problemet er
at hvis man fx. forestiller sig man har en cluster af Quad
processormaskiner, så er det vigtigt at kunne skelne imellem ting der
egner sig til at blive paralliseret lokalt, og dem der kan paralliseres
på en cluster, med den høje latenstid det medfører. Det kan være et
problem vis man prøver at "skjule" en cluster, således at folk lokkes
til at gøre ting der virker ineffektivt over et netværk.

Ups, maybe I should have written in english, should Jacob's letter have
attracted som Beowolf people.

Mvh Morten


 
Home   Subscribe   Mail Archive   Index   Calendar   Search

 
 
Questions about the web-pages to <www_admin>. Last modified 2005-08-10, 20:34 CEST [an error occurred while processing this directive]
This page is maintained by [an error occurred while processing this directive]MHonArc [an error occurred while processing this directive] # [an error occurred while processing this directive] *